1 COMP9318: Data Warehousing and Data Mining — L1: Introduction —
1
COMP9318: Data Warehousing and Data Mining
— L1: Introduction —
2
Chapter 1. Introduction
n Motivation: Why data mining?
n What is data mining?
n Data Mining: On what kind of data?
n Data mining functionality
n Are all the patterns interesting?
n Classification of data mining systems
n Major issues in data mining
3
Necessity Is the Mother of Inventionn Data explosion problem
n Automated data collection tools and mature database technology lead to tremendous amounts of data accumulated and/or to be analyzed in databases, data warehouses, and other information repositories
n We are drowning in data, but starving for knowledge!
n Solution: Data warehousing and data mining
n Data warehousing and on-line analytical processing
n Mining interesting knowledge (rules, regularities, patterns, constraints) from data in large databases
Who could be expected to digest millions of records, each having tens or hundreds of fields?
4
Evolution of Database Technologyn 1960s:
n Data collection, database creation, IMS and network DBMSn 1970s:
n Relational data model, relational DBMS implementationn 1980s:
n RDBMS, advanced data models (extended-relational, OO, deductive, etc.) n Application-oriented DBMS (spatial, scientific, engineering, etc.)
n 1990s: n Data mining, data warehousing, multimedia databases, and Web
databasesn 2000s
n Stream data management and miningn Data mining with a variety of applicationsn Web technology and global information systems
5
What Is Data Mining?
n Data mining (knowledge discovery from data)
n Extraction of interesting (non-trivial, implicit, previously
unknown and potentially useful) patterns or knowledge from
huge amount of data
n Data mining: a misnomer?
n Alternative names
n Knowledge discovery (mining) in databases (KDD), knowledge
extraction, data/pattern analysis, data archeology, data
dredging, information harvesting, business intelligence, etc.
n Watch out: Is everything “data mining”?
n (Deductive) query processing.
n Expert systems or small ML/statistical programs
6
Why Data Mining?—Potential Applications
n Data analysis and decision supportn Market analysis and management
n Target marketing, customer relationship management (CRM), market basket analysis, cross selling, market segmentation
n Risk analysis and management
n Forecasting, customer retention, improved underwriting, quality control, competitive analysis
n Fraud detection and detection of unusual patterns (outliers)
n Other Applicationsn Text mining (news group, email, documents) and Web miningn Stream data miningn DNA and bio-data analysis
7
Market Analysis and Management
n Where does the data come from?
n Credit card transactions, loyalty cards, discount coupons, customer complaint calls, plus (public) lifestyle studies
n Target marketing
n Find clusters of “model” customers who share the same characteristics: interest, income level, spending habits, etc.
n Determine customer purchasing patterns over time
n Cross-market analysis
n Associations/co-relations between product sales, & prediction based on such association
n Customer profiling
n What types of customers buy what products (clustering or classification)
n Customer requirement analysis
n identifying the best products for different customers
n predict what factors will attract new customers
n Provision of summary information
n multidimensional summary reports
n statistical summary information (data central tendency and variation)
8
Corporate Analysis & Risk Management
n Finance planning and asset evaluationn cash flow analysis and predictionn contingent claim analysis to evaluate assets n cross-sectional and time series analysis (financial-ratio, trend
analysis, etc.)n Resource planning
n summarize and compare the resources and spendingn Competition
n monitor competitors and market directions n group customers into classes and a class-based pricing proceduren set pricing strategy in a highly competitive market
9
Fraud Detection & Mining Unusual Patterns
n Approaches: Clustering & model construction for frauds, outlier analysisn Applications: Health care, retail, credit card service, telecomm.
n Auto insurance: ring of collisions n Money laundering: suspicious monetary transactions n Medical insurance
n Professional patients, ring of doctors, and ring of referencesn Unnecessary or correlated screening tests
n Telecommunications: phone-call fraudn Phone call model: destination of the call, duration, time of day or
week. Analyze patterns that deviate from an expected normn Retail industry
n Analysts estimate that 38% of retail shrink is due to dishonest employees
n Anti-terrorism
10
Other Applications
n Sports
n IBM Advanced Scout analyzed NBA game statistics (shots blocked,
assists, and fouls) to gain competitive advantage for New York
Knicks and Miami Heat
n Astronomy
n JPL and the Palomar Observatory discovered 22 quasars with the
help of data mining
n Internet Web Surf-Aid
n IBM Surf-Aid applies data mining algorithms to Web access logs
for market-related pages to discover customer preference and
behavior pages, analyzing effectiveness of Web marketing,
improving Web site organization, etc.
11
Data Mining: A KDD Process
n Data mining—core of knowledge discovery process
Data Cleaning
Data Integration
Databases
Data Warehouse
Task-relevant Data
Selection
Data Mining
Pattern Evaluation
12
Steps of a KDD Process
n Learning the application domainn relevant prior knowledge and goals of application
n Creating a target data set: data selectionn Data cleaning and preprocessing: (may take 60% of effort!)n Data reduction and transformation
n Find useful features, dimensionality/variable reduction, invariant representation.
n Choosing functions of data mining n summarization, classification, regression, association, clustering.
n Choosing the mining algorithm(s)n Data mining: search for patterns of interestn Pattern evaluation and knowledge presentation
n visualization, transformation, removing redundant patterns, etc.n Use of discovered knowledge
13
Data Mining and Business IntelligenceIncreasing potentialto supportbusiness decisions End User
BusinessAnalyst
DataAnalyst
DBA
MakingDecisions
Data PresentationVisualization Techniques
Data MiningInformation Discovery
Data Exploration
OLAP, MDA
Statistical Analysis, Querying and Reporting
Data Warehouses / Data Marts
Data SourcesPaper, Files, Information Providers, Database Systems, OLTP
14
Architecture: Typical Data Mining System
Data Warehouse
Data cleaning & data integration Filtering
Databases
Database or data warehouse server
Data mining engine
Pattern evaluation
Graphical user interface
Knowledge-base
15
Data Mining: On What Kinds of Data?
n Relational databasen Data warehousen Transactional databasen Advanced database and information repository
n Object-relational databasen Spatial and temporal datan Time-series data n Stream datan Multimedia databasen Heterogeneous and legacy databasen Text databases & WWW
16
Data Mining Functionalities
n Concept description: Characterization and discriminationn Generalize, summarize, and contrast data characteristics, e.g., dry
vs. wet regions
n Association (correlation and causality)n Diaper à Beer [0.5%, 75%]
n Classification and Predictionn Construct models (functions) that describe and distinguish classes
or concepts for future predictionn E.g., classify countries based on climate, or classify cars based
on gas mileagen Presentation: decision-tree, classification rule, neural networkn Predict some unknown or missing numerical values
17
Data Mining Functionalities (2)
n Cluster analysisn Class label is unknown: Group data to form new classes, e.g.,
cluster houses to find distribution patternsn Maximizing intra-class similarity & minimizing interclass similarity
n Outlier analysisn Outlier: a data object that does not comply with the general
behavior of the datan Noise or exception? No! useful in fraud detection, rare events
analysisn Trend and evolution analysis
n Trend and deviation: regression analysisn Sequential pattern mining, periodicity analysisn Similarity-based analysis
n Other pattern-directed or statistical analyses
18
Are All the “Discovered” Patterns Interesting?
n Data mining may generate thousands of patterns: Not all of them are interestingn Suggested approach: Human-centered, query-based, focused mining
n Interestingness measuresn A pattern is interesting if it is easily understood by humans, valid on new
or test data with some degree of certainty, potentially useful, novel, or validates some hypothesis that a user seeks to confirm
n Objective vs. subjective interestingness measuresn Objective: based on statistics and structures of patterns, e.g., support,
confidence, etc.n Subjective: based on user’s belief in the data, e.g., unexpectedness,
novelty, actionability, etc.
19
Can We Find All and Only Interesting Patterns?
n Find all the interesting patterns: Completeness
n Can a data mining system find all the interesting patterns?
n Heuristic vs. exhaustive search
n Association vs. classification vs. clustering
n Search for only interesting patterns: An optimization problem
n Can a data mining system find only the interesting patterns?
n Approaches
n First generate all the patterns and then filter out the uninteresting ones.
n Generate only the interesting patterns—mining query optimization
20
Data Mining: Confluence of Multiple Disciplines
Data Mining
Database Systems Statistics
OtherDisciplines
Algorithm
MachineLearning Visualization
21
Data Mining: Classification Schemes
n General functionalityn Descriptive data mining n Predictive data mining
n Different views, different classificationsn Kinds of data to be minedn Kinds of knowledge to be discoveredn Kinds of techniques utilizedn Kinds of applications adapted
22
Multi-Dimensional View of Data Miningn Data to be mined
n Relational, data warehouse, transactional, stream, object-oriented/relational, active, spatial, time-series, text, multi-media, heterogeneous, legacy, WWW
n Knowledge to be minedn Characterization, discrimination, association, classification,
clustering, trend/deviation, outlier analysis, etc.n Multiple/integrated functions and mining at multiple levels
n Techniques utilizedn Database-oriented, data warehouse (OLAP), machine learning,
statistics, visualization, etc.n Applications adapted
n Retail, telecommunication, banking, fraud analysis, bio-data mining, stock market analysis, Web mining, etc.
23
Major Issues in Data Miningn Mining methodology
n Mining different kinds of knowledge from diverse data types, e.g., bio, stream, Web
n Performance: efficiency, effectiveness, and scalabilityn Pattern evaluation: the interestingness problemn Incorporation of background knowledgen Handling noise and incomplete datan Parallel, distributed and incremental mining methodsn Integration of the discovered knowledge with existing one: knowledge fusion
n User interactionn Data mining query languages and ad-hoc miningn Expression and visualization of data mining resultsn Interactive mining of knowledge at multiple levels of abstraction
n Applications and social impactsn Domain-specific data mining & invisible data miningn Protection of data security, integrity, and privacy
24
Summary
n Data mining: discovering interesting patterns from large amounts of data
n A natural evolution of database technology, in great demand, with wide applications
n A KDD process includes data cleaning, data integration, data selection, transformation, data mining, pattern evaluation, and knowledge presentation
n Mining can be performed in a variety of information repositoriesn Data mining functionalities: characterization, discrimination,
association, classification, clustering, outlier and trend analysis, etc.n Data mining systems and architecturesn Major issues in data mining
25
A Brief History of Data Mining Society
n 1989 IJCAI Workshop on Knowledge Discovery in Databases (Piatetsky-
Shapiro)
n Knowledge Discovery in Databases (G. Piatetsky-Shapiro and W. Frawley, 1991)
n 1991-1994 Workshops on Knowledge Discovery in Databases
n Advances in Knowledge Discovery and Data Mining (U. Fayyad, G. Piatetsky-Shapiro, P. Smyth,
and R. Uthurusamy, 1996)
n 1995-1998 International Conferences on Knowledge Discovery in Databases
and Data Mining (KDD’95-98)
n Journal of Data Mining and Knowledge Discovery (1997)
n 1998 ACM SIGKDD, SIGKDD’1999-2001 conferences, and SIGKDD
Explorations
n More conferences on data mining
n PAKDD (1997), PKDD (1997), SIAM-Data Mining (2001), (IEEE) ICDM (2001), etc.
26
Where to Find References?n Data mining and KDD
n Conferences: ACM-SIGKDD, IEEE-ICDM, SIAM-DM, PKDD, PAKDD, etc.n Journal: Data Mining and Knowledge Discovery, KDD Explorations
n Database systemsn Conferences: ACM-SIGMOD, ACM-PODS, VLDB, IEEE-ICDE, EDBT, ICDT, DASFAAn Journals: ACM-TODS, IEEE-TKDE, JIIS, J. ACM, VLDBJ, etc.
n AI & Machine Learningn Conferences: Machine learning (ML), AAAI, IJCAI, COLT (Learning Theory), etc.n Journals: Machine Learning, Artificial Intelligence, etc.
n Statisticsn Conferences: Joint Stat. Meeting, etc.n Journals: Annals of statistics, etc.
n Visualizationn Conference proceedings: CHI, ACM-SIGGraph, etc.n Journals: IEEE Trans. visualization and computer graphics, etc.
Web resources:1. DBLP2. Google3. Citeseer4. DL@lib
27
Recommended Reference Booksn I. H. Witten and E. Frank, Data Mining: Practical Machine Learning Tools and Techniques with Java
Implementations, Morgan Kaufmann, 2001
n C. C. Aggarwal, Data Mining: The Textbook, Springer, 2015��
n J. Leskovec, A. Rajaraman, and J. Ullman, Mining of Massive Datasets (v2.1), Cambridge University Press, 2014.
n Y. S. Abu-Mostafa, M. Magdon-Ismail, and H.-T. Lin, Learning From Data. AMLBook, 2012.
n J. Han and M. Kamber. Data Mining: Concepts and Techniques. Morgan Kaufmann, 2001
n D. J. Hand, H. Mannila, and P. Smyth, Principles of Data Mining, MIT Press, 2001
n T. Hastie, R. Tibshirani, and J. Friedman, The Elements of Statistical Learning: Data Mining, Inference, and Prediction, Springer-Verlag, 2001
n T. M. Mitchell, Machine Learning, McGraw Hill, 1997
n P-N. Tan, M. Steinbach, and V. Kumar, Introduction to Data Mining,. Addison-Wesley, 2005
n S. M. Weiss and N. Indurkhya, Predictive Data Mining, Morgan Kaufmann, 1998
Jai’s Project (COMP9318, 2016s2)
n Problemn http://kentandlime.com.au/, a startup company helping
male customers to stay in fashion but out of the shops.n Status-quo:
n Ask questions, and stylists makes a list of recommended items, and send them to customers
n If happy, customers pay for the product. n Recommendation is the key!
n Challengesn Dirty datan Not an easy/typical recommendation system settingsn Customer feedbacksn Real-time recommendations
28
http://www.news.com.au/lifestyle/fashion/fashion-trends/fashions-most-unlikely-trend-would-you-buy-clothes-chosen-for-you/news-story/8634b5f06f608b9f2fd7c27758f9c10a
Solutions - Highlight
n Use domain-knowledge and quick evaluations to guide the whole process
n Data preprocessingn Data source: CRM (profile) + NoSQL DB (transactions)
n Missing data: e.g., due to schema changes
n Data normalization: A’s XL = B’s L
n Data noise: k-means / binning
n Data selection: remove sparse columns/rows
n Feature engineeringn weight-to-height ratio
29
Solutions – Highlight /2
n Product class clustering and predictionn Collaborative filtering with smoothing and
weightingn Content-based recommendation (solve the cold
start problem)n Incorporate customer feedbacksn Association rule mining:
n LSShirts_1, Shorts_2 è Socks_3n Emsemble of the above
n Plus many engineering efforts
30
Results
n Test set: n Classification rate: 74%, on par with humans
n Deployed to production on 18-24 Nov 2016:n Customers rejecting on average 2.36 items out of a
basket of 10-12 items è (76.4%, 80.3%)n Latency: 2.3s
n Future work identifiedn e.g., seasonality
n Check Jai’s presentation slides for more details.
31